The end of the 2026 Semiconductor Cycle

Reading the Samsung Tealeaves

The Samsung Semiconductor Memory story is an amazing one. From humble beginnings in the 1990s, when Samsung had to pay a penalty after a memory purchaser bought a faulty rice cooker from the company and suspected the memory had the same quality, the Koreans snatched the lead from the Americans.

Most US memory companies folded while Samsung began to rival Intel as the most important semiconductor company ever. With a 40% share over a decade, Samsung seemed untouchable, and plans were underway to solidify its position.

In early 2021, development of PIM (Processing In Memory) began, building on HBM (High Bandwidth Memory) technology, and in 2022, the first demonstration was unveiled on an AMD GPU.

Long story short - Samsung fumbled the ball on HBM. Central to the story were several misses in SK Hynix's qualifications for HBM3E, but the problems at Samsung extended beyond HBM and affected broader operations.

In the horror year 2024, Samsung went from over 60% market share in HBM to less than 20%, while SK Hynix became Nvidia's preferred HBM supplier.

But there is light at the end of the tunnel. The overall Q3-25 result was the best in 3 years and is getting close to the last peak.

Not only did the revenue recover, but profits jumped, with net profits improving by more than 100% YoY. Even though Samsung is a conglomerate, the company’s memory business is crucial to profitability and is subject to the broader memory cycle.

Samsung's complex business model has surprising elements that give the conglomerate certain advantages. The choppy mobile phone business, quite different from Apple’s Q4-centric business, is neatly counterbalanced by the display and other divisions, providing a relatively stable revenue base.

Samsung's significant cyclical trend aligns with memory revenue and, by extension, the broader memory cycle. This is also what is driving the increase in Q3-25 revenue.

The dominance is even more significant when analysing operating profits by division.

While the other divisions deliver relatively steady operational profits, the Memory division is either a high-performance engine in the Samsung Supertanker or it is a massive anchor. In the last up cycle, the Memory business temporarily delivered more operating profits than all the other divisions combined, and the course is set for even greater reliance on the Memory division in the next few quarters.

While the jump in operating profits of the memory division could be due to increased production, this is not entirely the case for Samsung.

Samsung has, for a long time, been running at full capacity, irrespective of whether production has been golden or has created massive losses. This is the “Thelma and Louise” operating model of the large memory companies. Full speed over the cliff.

While production is up, most of the revenue gain comes from steep price increases in the market.

The inventory position confirms this.

Despite a decline in finished goods inventory, it is not sufficient to account for the sharp increase in revenue. Samsung's current performance increase is mainly linked to spiking memory prices.

What is worrying from a supply perspective is that current production capacity is insufficient to sustain the current sales rate, as finished goods inventories continue to decline.

Adding to the concern is that the Work-In-Progress inventory has flatlined, as this is a proxy for production activity levels. There is no indication that Samsung is close to a meaningful increase in production capacity.

While the memory production of Samsung has been steadily increasing from a GB perspective as the technological development has improved the memory cell size, but like all the other memory companies, the Koreans got scared by the depth of the last cycle and have held back on investments.

This is confirmed by analysing Semiconductor CapEx relative to the maintenance CapEx required to keep the manufacturing assets at the same level. In the downcycle, the Company balanced investments at about the maintenance level, but over the last three quarters, the investment level has fallen to a very low level. In Q3, Samsung should have invested $9.6B in maintenance before any additional capacity investments, but the Koreans dished out a meagre $5.6B.
This is far from sufficient for a meaningful increase in manufacturing capacity.